Workloads in the cloud and workloads on premises are both reasonable choices. Enterprises are about evenly split between the two.
Server huggers and cloud bigots of the world rejoice! You’re both right. Unfortunately, you’re also both wrong. “After 18 years of public cloud, we’ve had a long time to sort out which apps go where,” writes VMware’s Michael Coté, and we seem to have landed on a 50:50 equilibrium, with roughly half of workloads sticking with on-premises data centers, and the other half going to public cloud.
This wasn’t how the cloud computing narrative was supposed to go. Nearly 10 years ago, for example, Gartner called out how much faster VM growth was in public cloud environments than private clouds. By 2021 cloud remained top of mind for every enterprise, but cloud spend was still a relative pittance compared to on-premises IT. Was the cloud party a lie?
No, of course not, as the big three cloud companies’ recent earnings calls all highlighted. AWS, Microsoft, and Google all saw their respective cloud businesses increase their pace of growth, with AI fueling a significant chunk of that growth. And yet, as I’ve written recently, “it depends” tends to be the right guiding principle for enterprise IT, but this time it’s because of the people involved and not the technology.
Hope and hype in cloud land
My InfoWorld colleague David Linthicum may be overstating the case against cloud when he argues that “cost, loss of control, lack of return on investment, data privacy, and even outright misleading marketing” have a swelling population of enterprises hitting the brakes on their cloud plans. Still, he has a point. For some time, the cloud carried a certain aura that resisted close inspection of its claims. Today, he says, enterprise second thoughts on cloud “can potentially drive a resurgence of cloud services that are better attuned to the intricate demands of enterprise IT.”
The idea was always that the enterprise would change to be more like the cloud. Instead, we may see the cloud change to be more like the enterprise. Or a happy compromise between the two extreme positions.
What we’re not seeing is a big pendulum swing back to on-premises deployments. Linthicum suggested that 2023 might be the year of cloud repatriation. It wasn’t. Instead we saw AI drive more spend to the cloud. I suspect that a significant part of any inclination to repatriate cloud applications back to private data centers is a failure to refactor applications for cloud in the first place, which tends to be the reason for the cloud costing more. Lifting and shifting applications to the cloud without refactoring those applications to take advantage of cloud architecture and services makes it virtually impossible for the cloud to cost less. But AI applications aren’t lift-and-shift workloads; they’re built for the cloud.
Unfortunately, many are also a complete waste of money, as I’ve indicated, which brings us back to cloud equilibrium.
Where are all those workloads going?
On any given day, it’s anyone’s guess where the next set of workloads will go, as a recent Barclays CIO survey suggests (shared by Michael Dell here):
Barclays CIO Survey
This seems to point to more repatriation, but really it’s just plans to repatriate. The actual repatriation is almost certainly much less. Just as we saw plans to move everything to the cloud back in the heady early days of cloud, we’re likely to see plans to move a significant chunk of applications back to private data centers fizzle out, too. Why? Because moving applications is hard.
As such, rather than measuring plans, it’s worth checking on where workloads actually go. According to that same Barclays CIO survey, the answer is cloud.
Barclays CIO Survey
That’s a lot of cloud movement, but with all that interest in repatriation, it may simply signal that, as Coté posits, we’re due to settle into prolonged equilibrium, with plenty of workloads moving to the cloud but many staying on premises. Underlying it all may be that enterprises don’t have the right people to refactor applications for the cloud. Or, rather, that the teams that got them where they are may not be the right teams to take them where they need to go. That’s a people problem, as most technology issues are. Change will take decades.
This is why, for most workloads, the answer to whether it goes to the cloud or stays on premises is, “It depends.” That is, it depends on the people involved, and not necessarily on the workload itself.
